XLMiner^® Overview

XLMiner is a comprehensive, fully integrated data mining add-in for Microsoft^® Excel developed by Cytel Inc..

Topics and methods available in the XLMiner tool include classification (classification trees, discriminant analysis, logistic regression, neural networks, naive Bayes, k-nearest neighbors), prediction (multiple linear regression with subset selection, k-nearest neighbors, regression trees), affinity analysis (association rules), data reduction and exploration (principal components, k-means clustering, hierarchical clustering), and data preparation (partitioning into training, validation and test data sets, stratified sampling for rare outcomes, missing data imputation, binning, transformation of categorical variables into dummies or category scores). XLMiner also extends Excel's graphics with matrix scatter plots, dendrograms, box plots, and lift charts. Because XLMiner is Excel-based, it has a short learning curve for those familiar with Excel. It can also integrate easily with other Excel-based analysis the user may already be doing.

XLMiner uses Excel primarily as an interface and platform - XLMiner's data mining modules do not depend on Excel's statistical calculations. While Excel limits the user to 60,000 rows, XLMiner allows the user to sample from a much larger database, do the analysis on a statistically-valid sample, and, for supervised learning, score the results back out to the database.

XLMiner's education version was released in 2002 and is in use at MIT's Sloan School, Duke's Fuqua School, Georgetown's McDonough School, the University of Maryland's Smith School, Indiana University, Boston University, and elsewhere. The standard edition of XLMiner was released in 2003.

A review of XLMiner, along with product comparisons to SAS Enterprise Miner, SPSS Clementine, GhostMiner, and Quadstone, appears in the November, 2003, issue of The American Statistician.